ABSTRACT Linking experience and movement is fundamental to nearly all life, as is the ability to predict, based on previous experience, that a condition warrants a motor response. Such associations often rely on the cerebellum, which harnesses diverse sensory, motor, and cognitive information encoded by a densely-packed layer of granule cells (GC) as a substrate for learning. Classical models of cerebellar learning posit that synaptic inhibition of GCs from local interneurons called Golgi cells is essential for establishing a non-overlapping, sparse population code. However, recent studies have called such models into question by revealing that unimodal sensory stimulation can activate widespread GC activity, suggesting that responses can be dense and redundant even for simple sensory representations under some conditions. These observations call into question whether and how GCs encode discrete stimulus features at the population level and the degree to which synaptic inhibition generates sparsity and pattern separation. To test how GCs encode stimulus features at the population level and how these representations are regulated by local inhibition, I propose to measure stimulus evoked GC activity using video-rate in vivo calcium imaging in awake mice in conjunction with a GC-specific manipulation of synaptic inhibition. In Aim 1, I will test whether and how GCs encode feature-specific representations of stimulus intensity and identity by measuring how single and population GC calcium transients change in response to varied stimulus features over many randomly interleaved trials, while controlling for movement- related activity with high-speed video and vibration sensors. Establishing when and how stimulus features are represented by unique or overlapping GC populations will allow us to test how these activity patterns are shaped by local inhibition. In Aim 2, I will test how GCs integrate multiple sensory inputs. I will image GC responses to combined sensory stimuli of either different modalities (i.e. auditory and somatosensory) or the same modality (i.e. two tones with differing frequencies) and measure how single and population GC calcium transients change in response to additional, simultaneous stimuli. Based on preliminary data, I anticipate that, while converging inputs can be summed to drive GCs past spike threshold, GCs will exhibit subtractive, as well as additive, signal integration due to stimulus-specific recruitment of inhibition. Finally in Aim 3, I will test how inhibition from Golgi cells shapes the population responses identified in Aim 1 and enables the multisensory integration by GCs assessed in Aim 2 by acutely blocking GABAergic inputs to GCs using a novel enzymatic capture system. These experiments will provide novel insight into how GCs encode information at the individual and population level and reveal how this encoding is regulated by local synaptic inhibition. Hence, this study will test longstanding models of cerebellar function by establishing how the GC layer represents patterns of sensory information in a manner that enables associative learning.